Prenatal Intake of High Multivitamins or Folic Acid With or Without Choline Contributes to Gut Microbiota‐Associated Dysregulation of Serotonin in Offspring

ABSTRACT

The gut microbiota is amenable to early nutrition including micronutrients but intake above and below the recommendations commonly occur with unknown consequences. Serotonin (5‐hydroxytryptamine [5‐HT]) is a monoamine found centrally and peripherally with diverse functions such as food intake regulation via the hypothalamic 5‐HT receptor 2C (5‐HTR2C). This study determined the impact of prenatal micronutrients on the gut microbiota and serotonergic system in offspring. Pregnant Wistar rats were fed either recommended vitamins (RV), high vitamins (HV), high folic acid with recommended choline (HFRC), or high folic acid with no choline (HFNC). Offspring were fed a high‐fat diet for 12 weeks postweaning. HV, HFRC, and HFNC males and females had lower hypothalamic 5‐HTR2C protein expression compared to RV. Brain 5‐HT concentrations were lower but colon 5‐HT concentrations were higher in HV and HFNC males and females and HFRC males compared to RV. Refeeding response after 5‐HTR2C agonist was negatively correlated with hypothalamic 5‐HTR2C protein expression in males and with brain 5‐HT concentrations in females. Random forest revealed top bacterial taxa, which Lactococcus, Ruminococcus, Bacteroides, and Oscillospira showed significant correlations with refeeding response and concentrations of brain and colon 5‐HT. In conclusion, excess or imbalanced prenatal consumption of micronutrients leads to gut microbiota‐associated disturbances in the serotonergic system in offspring.

Consuming too much vitamins including folic acid or not enough choline during pregnancy can disturb the gut–brain serotonergic system in adult offspring. Lactococcus, Ruminococcus, Bacteroides, and Oscillospira were associated with altered serotonergic expression and function in male and female offspring, highlighting unique gut microbiota patterns that occur with varying micronutrient composition.

Abbreviations

5‐HT

5‐hydroxytryptamine

5‐HTR2C

5‐HT receptor 2C

ANOVA

one‐way analysis of variance

HFNC

high folic acid with no choline

HFRC

high folic acid with recommended choline

HV

high vitamins

mCPP

m‐chlorophenylpiperazine

PCoA

principal coordinates analysis

PERMANOVA

permutational multivariate analysis of variance

QIIME2

Quantitative Insights Into Microbial Ecology 2

RV

recommended vitamins

TBST

Tris‐Buffered Saline with Tween 20

1. Introduction

The gastrointestinal tract harbors a complex array of microorganisms that regulates the host production of signaling molecules [1]. Serotonin (5‐hydroxytryptamine [5‐HT]) is a multifunctional monoamine best known as a neurotransmitter in the central nervous system for the modulation of appetite, mood, memory, and reward but extends to other processes including cardiovascular function, respiratory drive, pain processing, digestion, and immune response [2]. There are at least 15 receptor subtypes in the 5‐HT family [3], and among them, the 5‐HT receptor 2C (5‐HTR2C) has been found to largely mediate the appetite‐suppressing actions of 5‐HT in the hypothalamus [4]. Only a small portion of the total 5‐HT is found in the brain, whereas over 90% of 5‐HT is known to be produced in the gut [5]. The importance of the gut microbiota in 5‐HT signaling has been demonstrated in studies using germ‐free mice, in which a lack of gut microbiota resulted in alterations in brain 5‐HT turnover and lower fecal, cecum, and colon 5‐HT concentrations compared to conventionally colonized controls [6, 7, 8]. Germ‐free mice also displayed aberrant anxiety‐like behaviors supporting an emerging link with depression and neurological disorders [9, 10, 11]. However, the relationship between the gut microbiota and 5‐HT in the context of metabolic phenotypes has received limited consideration.

Nutritional stimuli and changes in the gut microbiota are thought to shape long‐term metabolic outcomes as contributors to health and disease [12]. The composition of the gut microbiota can be modified by various external factors including diet, stress, and medications [13]. In North America, the prevalence of obesity has increased [14, 15] in tandem with micronutrient consumption above and below the recommendations during pregnancy [16, 17, 18]. Folic acid is commonly over‐consumed in a current environment of widely available dietary supplements and liberalized fortification policies [16, 17, 18]. Choline, an essential nutrient necessary for structural and regulatory functions in the body [19], in contrast is absent or found in minimal amounts in prenatal supplements and that a majority of the population is not meeting the Adequate Intake level for choline [16, 17, 18]. Despite these intake patterns, the impact on metabolic health of offspring and underlying mechanisms remain unclear.

This research builds upon our initial studies reporting that a prenatal intake of nontoxic 10x multivitamins leads to greater food intake, body weight, and characteristics of the metabolic syndrome in male and female offspring [20, 21]. When 10x folic acid with recommended choline was provided, the obesogenic phenotypes were observed in male offspring but not in female offspring [21, 22]. These contrasting responses were abolished with the removal of choline from 10x folic acid producing obesogenic phenotypes in both male and female offspring [21], suggesting choline as a potential modulator of the effects of folic acid. In our recent work, we found that colon 5‐HT concentrations were elevated with expression of the obesogenic phenotypes in offspring of dams fed an excessive or imbalanced amount of micronutrients prenatally, and were strongly associated with fasting blood glucose [23]. In addition to peripheral 5‐HT, we showed that the central serotonergic system appears to be responsive to prenatal micronutrient exposure with the use of 5‐HTR2C agonist, meta‐chlorophenylpiperazine (mCPP) [23]. Based on compelling evidence for host–microbiota interactions impacting 5‐HT‐related complex behaviors [8], we reasoned that the gut microbiota is an important determinant of the variations in responses to maternal micronutrient exposure. To date, no study has investigated central and gut 5‐HT concentrations concertedly with hypothalamic 5‐HTR2C expression to capture the overall disturbances in the serotonergic system and their alterations in relation to the gut microbiota composition.

Here, we set out to fill in critical gaps in our understanding of the role of prenatal micronutrient consumption in gut microbiota signatures and serotonergic system in offspring. The objective of this research was to determine 5‐HT concentrations, 5‐HTR2C expression and function, and association with the gut microbiota in adult offspring born to dams fed excess vitamins or folic acid with or without choline during pregnancy. We hypothesized that the serotoninergic system of offspring is altered centrally and peripherally due to micronutrient consumption above or below the recommendations in association with gut microbiota dysregulation.

2. Experimental Section

2.1. Study Design

First‐time pregnant (2–4 days gestation) Wistar rats (Charles River, Wilmington, MA, USA) were singly housed in a controlled room maintained at 22 ± 1°C under a 12:12‐h light–dark cycle (lights on at 0700). Pregnant dams (n = 10–12/group) were allocated randomly to either the control AIN‐93G diet (Research Diets, New Brunswick, NJ, USA) [24] containing recommended 1x vitamins (recommended vitamins [RV]) or modified AIN‐93G diet with either high 10x vitamins (high vitamins [HV]), high 10x folic acid with recommended 1x choline (high folic acid with recommended choline [HFRC]), or high 10x folic acid with no choline (high folic acid with no choline [HFNC]). At birth, each litter was culled to 10 offspring and dams were switched to the RV diet throughout lactation. From weaning until 12 weeks postweaning, one male and one female offspring from each litter were housed individually and fed a high‐fat AIN‐93G diet containing 60 kcal% fat, from mostly lard (Research Diets, New Brunswick, NJ, USA). At 6 weeks postweaning, rats were fasted overnight then injected intraperitoneally with either mCPP (2.5 mg/kg) or 0.9% saline in a crossover fashion with a 72‐h washout period in‐between. Food was provided immediately after the injection and food intake was measured for 1‐h, which refeeding response was expressed as the difference in intake amount in grams after mCPP and saline. All rats had ad libitum access to food and water throughout the experiment. At 12 weeks postweaning, all rats (one male and one female offspring from each litter; n = 10–12/group) were terminated by decapitation following an overnight fast. The whole brain was rapidly removed from the cranium, and the hypothalamus was dissected on a frozen glass plate rested on ice using anatomical landmarks from the Rat Brain Stereotaxic Coordinates as previously outlined [20]. The entire length of the colon was excised from which fecal samples were collected directly and the distal 0.5 cm segment was aliquoted separately. All samples were frozen on dry ice and then stored at −80°C until further use. The protocol was approved by the Institutional Animal Care and Use Committee (#10113) at Utah State University, where the animal experiments originated.

2.2. Justification of Diet and Dose Information

Table S1 shows diet composition. All diets were balanced for micronutrient‐to‐energy content. The AIN‐93G diet contains 2 mg of folic acid per kg diet in 4000 kcal, which is accepted as the basal dietary requirement for rodents [24]. This dose has been compared to the recommended daily allowance of 400 µg of dietary folate for women and calculated based on the Food and Drug Administration conversion factors for doses from humans to animals [25]. In HFRC and HFNC, the dose of 20 mg of folic acid per kg diet reflects 4000 µg folic acid per day in women. This dose was selected based on our prior work, known to be nontoxic, nonteratogenic, and results in obesogenic phenotypes, thus served as a reproducible benchmark [20, 22, 26]. The AIN‐93G diet contains 1 g free choline per kg diet and was removed in HFNC to isolate the effect of maximal imbalance between folic acid and choline. This dose was consistent with the observation that no folate‐related neural tube defects are known to occur under choline‐deficient conditions [27]. For the postweaning diet, we chose a high‐fat diet intended to accelerate the obesogenic phenotypes in offspring allowing for a shorter experimental timeframe compared to our previous studies that used a regular‐fat diet [20, 22].

2.3. Western Blotting

The hypothalamus samples were prepared using RIPA lysis buffer (#20‐188; Sigma–Aldrich, St. Louis, MO, USA) supplemented with protease and phosphatase inhibitor cocktail (#PI78441; Thermo Fisher Scientific, Wilmington, DE, USA). The homogenate was maintained shaking for 2 h at 4°C then centrifuged at 14 000 rpm for 20 min at 4°C. Protein concentration was determined using a bicinchoninic acid (BCA) assay kit (Pierce, Cambridge, NJ, USA) with bovine serum albumin as a standard. We employed stain‐free imaging technology, which uses a polyacrylamide gel containing a trihalogenic compound covalently crosslinked to tryptophan residues, whereby UV light excitation triggers fluorescence. This technology differs from traditional western blotting methods as it allows the trihalogenic‐modified proteins to be transferred to membranes, enabling a more reliable loading control from stain‐free total protein measurement than the use of housekeeping proteins [28]. Samples were mixed with 5 µL Laemmli sample buffer containing 5% β‐mercaptoethanol, boiled for 5 min and 32 µg/sample were separated using electrophoresis with the Mini‐PROTEAN 4%–15% gradient TGX Stain‐Free gel (BioRad, Hercules, CA, USA) at 200 V for 45 min. The stain‐free gel was activated for 45 s and imaged using the ChemiDoc Touch imaging system (BioRad, Hercules, CA, USA). The gel was transferred semidry to a polyvinylidene difluoride (PVDF) membrane using a Turbo transfer system (BioRad, Hercules, CA, USA) at 2.5 A for 7 min. The PVDF membrane was blocked with gentle agitation at room temperature for 1 h using Everyday Blocking Buffer (#12010020; BioRad Hercules, CA, USA). The 5‐HTR2C primary antibody (#SC‐17797, Santa Cruz Biotechnology, Dallas TX, USA; 1:1000) was diluted with Everyday Blocking Buffer and incubated overnight at 4°C. The blotting membrane was washed five times for 5 min with Tris‐Buffered Saline with Tween 20 (TBST; 20 mM Tris, 500 mM NaCl, pH 7.4, 0.05% v/v Tween 20) and incubated with horseradish peroxidase‐conjugated anti‐mouse IgG (#7076, Cell Signaling Technology, Danvers, MA, USA; 1:8000) diluted in TBST at room temperature for 1 h. The membrane was washed six times for 5 min with TSBT, and then total protein on the blot was detected using the ChemiDoc Touch imaging system. The enhanced substrate solution (Clarity Western ECL MAX, BioRad, Hercules, CA, USA) was applied for 5 min, and the chemiluminescent signals were captured with the ChemiDoc Touch imaging system. Band densitometry normalized to the total protein content was performed using the lane and band tool with the Image Lab software (BioRad, Hercules, CA, USA) [29].

2.4. Enzyme‐Linked Immunosorbent Assay

The whole brain and distal colon were used to determine 5‐HT concentrations. Tissue samples were finely ground at 4°C in prefilled bead mill tubes (15340154; Thermo Fisher Scientific, Wilmington, DE, USA) using a Bead Mill 24 Homogenizer (Thermo Fisher Scientific, Wilmington, DE, USA), weighed and homogenized in 0.9% NaCl phosphate‐buffered saline supplemented with 0.1% ascorbic acid at 50 mg/mL. Total protein was quantified using a BCA assay (Pierce, Cambridge, NJ, USA). An enzyme‐linked immunosorbent assay was performed in accordance with the manufacturer's protocol (SEJ39‐K01, Eagle Biosciences, Amherst, NH, USA) and adjusted for total protein content.

2.5. 16S rRNA Sequencing

Fecal DNA was extracted using the DNeasy PowerSoil Kit (Qiagen, Hilden, Germany) following the manufacturer's guidelines. The concentration was measured using the NanoQuant spectrophotometer (Tecan, Männedorf, Switzerland). The V4 hypervariable region was amplified using the forward and reverse primers, 16S‐515F and 16S‐806R, with unique barcoding designed for dual‐indexing [30]. PCR amplicons were confirmed by gel electrophoresis and subsequently purified and quantified. Pooled samples were stored at −20°C before sequencing on a MiSeq instrument (Illumina, San Diego, CA, USA) with a V2 500 cycle kit for paired‐end reads.

2.6. Statistical Analyses and Bioinformatics

Data analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) except in the case of microbiome analyses for beta diversity and subsequent modeling approaches, which were conducted in Quantitative Insights Into Microbial Ecology 2 (QIIME2) and MicrobiomeAnalyst [31]. Model assumptions were assessed using the Shapiro–Wilk test for normality and Levene's test for homogeneity of variance. A one‐way analysis of variance (ANOVA) by the PROC MIXED model procedure was used to analyze food intake response to mCPP; protein expression of 5‐HTR2C in the hypothalamus normalized by total protein; and 5‐HT concentrations in the brain and colon. In the event the effects for diet were significant, Tukey's post‐hoc test was performed. The relative abundance percentage of select genera was compared among the diet groups using the Kruskal–Wallis test by the PROC MIXED model procedure, which upon significant group effects were followed by Dunn's test.

The sequencing data were processed using the QIIME2 [32] version 2020.11. Raw sequences were demultiplexed and quality‐filtered using DADA2 to trim, denoise and checked for chimeras [33]. The MAFFT plugin [34] was used for sequence alignment, and taxonomy was assigned against the latest version of the SILVA 138 database at 99% similarity [35]. Downstream analyses were conducted within the Marker Data Profiling module of MicrobiomeAnalyst 2.0 [31]. Features that are low abundance may be less functionally important thus the low count filter of 10% across samples was applied. Moreover, features that are of low variance may be less informative for comparative analysis thus low variance filter at 10% based on the interquartile range was applied. Beta diversity was determined based on the Jaccard index distance at the genus level and visualized with a principal coordinates analysis (PCoA) ordination. A permutational multivariate analysis of variance (PERMANOVA) was used to assess differences in beta diversity with corrections for multiple testing based on the Benjamini–Hochberg procedure. Random forest modeling [36] was implemented within Microbiomeanalyst [31] that uses a supervised machine learning algorithm to provide estimates of relative importance in which bacterial taxa is predictive of classification by group. The number of trees to be used for classification was set to 500 and the number of predictors for each node at 7. Beta diversity and random forest modeling required aggregated data sets for male and female offspring for the determination of the most informative attributes of the entire microbiome data. High abundance was shown in red, and low abundance was shown in blue using a heatmap scale to indicate differential abundance in the top genera. The Spearman correlation coefficient was used to determine the association among outcome measures (5‐HTR2C expression, 5‐HT concentrations, and relative abundance of bacterial taxa) and was represented in a matrix to depict the strength and direction of the relationship.

One offspring per sex per litter represented an independent statistical unit that addresses the issues of design and analysis with the use of multiparous species in developmental nutritional studies [37]. Our sample size included n = 8–9/group for 5‐HTR2C, n = 8–12/group for refeeding response, n = 9–12/group for 5‐HT measurements, and n = 10–11/group for gut microbiota analyses, due to limited sample availability or failures in injection and extraction techniques. The significance level was set at p ≤ 0.05 and trending at 0.05 < p ≤ 0.1. Values are presented as mean ± standard error of the mean (SEM) for all data except for the abundance of taxa represented as relative percentages.

3. RESULTS

3.1. Hypothalamic 5‐HTR2C Protein Expression and Function

In male and female offspring, hypothalamic protein expression of 5‐HTR2C was lower with prenatal exposure to HV, HFRC, and HFNC relative to the respective RV group (p < 0.0001 for both male and female offspring; Figure 1A, B). The function of 5‐HTR2C was assessed with food intake 1‐h after the injection with the receptor agonist mCPP, which is known to induce acute hypophagia [38] versus saline. The refeeding response was higher in HV males compared to the control offspring (p < 0.05; Figure 1C). HFRC males were not different from HV nor RV whereas HFNC males were not different from RV but had lower food intake response than HV males. Female offspring from HFRC‐fed dams did not differ from the control group but had lower refeeding response compared to HV or HFNC females (p < 0.01; Figure 1D).

FIGURE 1.

Protein expression of 5‐hydroxytryptamine 2C receptor (5‐HTR2C) in the hypothalamus and refeeding response 1‐h after the intraperitoneal injection with 5‐HTR2C receptor agonist m‐chlorophenylpiperazine (mCPP) in male and female offspring exposed to varying micronutrients prenatally (A–D). Blots were produced using stain‐free imaging technology, and band densitometry was normalized to the total protein content. Refeeding response was expressed as the difference in food intake amount in grams after mCPP and that after saline. Diets are abbreviated as HFNC, high 10x folic acid (20 mg/kg diet) with no choline; HFRC, high 10x folic acid (20 mg/kg diet) with recommended choline (1 g free choline/kg diet); HV, high 10x vitamins; RV, recommended vitamins. Distinct superscripts show p ≤ 0.05 determined by one‐way ANOVA using the PROC MIXED procedure, which any significant group effect was followed by Tukey's post‐hoc test. Values are shown as mean ± SEM; n = 8–9/group for 5‐HTR2C and n = 8–12/group for refeeding response with independent experiments performed in singlicate measurements. ANOVA, analysis of variance; SEM, standard error of the mean.

3.2. Brain and Colon 5‐HT Concentrations

Male offspring of HV, HFRC, or HFNC‐fed dams had lower brain 5‐HT concentrations compared to RV (p < 0.0001; Figure 2A). Female offspring of HV or HFNC‐fed dams had lower brain 5‐HT concentrations compared to RV, with no differences in brain 5‐HT between HFRC and RV females (p < 0.0001; Figure 2B). Colon 5‐HT concentrations were higher in HV, HFRC, and HFNC males compared to the control offspring (p < 0.01; Figure 2A). HV or HFNC females had higher colon 5‐HT concentrations compared to RV, with no differences between HFRC and RV females (p < 0.001; Figure 2B).

FIGURE 2.

Tissue concentrations of 5‐hydroxytryptamine (5‐HT) in the brain and colon (ng/mg) in male and female offspring exposed to varying micronutrients prenatally (A, B). Diets are abbreviated as HFNC, high 10x folic acid (20 mg/kg diet) with no choline; HFRC, high 10x folic acid (20 mg/kg diet) with recommended choline (1 g free choline/kg diet); HV, high 10x vitamins; RV, recommended vitamins. Distinct superscripts show p ≤ 0.05 by one‐way ANOVA using the PROC MIXED procedure, which any significant group effect was followed by Tukey's post‐hoc test. Values are shown as mean ± SEM; n = 9–12/group with independent experiments performed in technical duplicate measurements. ANOVA, analysis of variance; SEM, standard error of the mean.

3.3. Gut Microbiota Composition

The gut microbiota is recognized to have an important role in host health [39], thus we employed 16S rRNA sequencing to determine the gut microbiota composition in offspring exposed to varying amounts of micronutrients. In the PCoA, 45% of the variation was explained by the first principal coordinate. Beta diversity using the Jaccard index distance was significantly different at the genus level (PERMANOVA p = 0.004; Figure 3). We leveraged random forest analysis to identify the most important variables in the model based on 500 decision trees with a decreasing order of mean decrease accuracy (Figure 4). The mean out of bucket estimate was 0.40. Among the top 12 genera, those with the mean decrease accuracy score above 0.002 were Adlercreutzia, Lactococcus, Ruminococcus (from Ruminococcaceae), Bacteroides, Lactobacillus, and Oscillospira with overall differential abundance expressed using a heatmap, illustrating that prenatal diet was an important determinant of the gut microbiota signature in offspring. Building upon differences in the gut microbiota composition from random forest modeling, we investigated the relative abundance of bacterial taxa (Figure 5A, B). In male offspring, HV, HFRC, and HFNC did not differ in Bacteroides from RV but the HFRC group had higher abundance compared to HV males (p < 0.05). Oscillospira was lower in HFRC males compared to the RV control with HV that did not differ from all other groups (p < 0.01). Lactococcus and Lactobacillus were lower in HFNC males when compared to HV but not with RV, with HFRC not different from HV and HFNC (p < 0.01 for Lactococcus, p < 0.05 for Lactobacillus). In female offspring, HV, HFRC, and HFNC did not differ in Adlercreutzia compared to RV, but the abundance was lower in HFNC compared to HV, with HFRC being between HV and HFNC (p < 0.05).

FIGURE 3.

Principal coordinates analysis of the Jaccard index distance metrics for pooled datasets of male and female offspring exposed to varying micronutrients prenatally. The principal coordinate (PC) axis shows the percentage of variations explained by the data. Diets are abbreviated as HFNC, high 10x folic acid (20 mg/kg diet) with no choline; HFRC, high 10x folic acid (20 mg/kg diet) with recommended choline (1 g free choline/kg diet); HV, high 10x vitamins; RV, recommended vitamins. PERMANOVA with corrections for multiple testing p = 0.004; n = 10–11/group with independent experiments performed in singlicate measurements. PERMANOVA, permutational multivariate analysis of variance.

FIGURE 4.

Top genera identified by random forest modeling with predictive power for classification among the diet groups. The mean decrease accuracy values quantify the importance of the variable for prediction with the genera shown in a decreasing order. The number of trees to be used for classification was set to 500 and the number of predictors for each node at 7. Diets are abbreviated as HFNC, high 10x folic acid (20 mg/kg diet) with no choline; HFRC, high 10x folic acid (20 mg/kg diet) with recommended choline (1 g free choline/kg diet); HV, high 10x vitamins; RV, recommended vitamins. Heatmap scale shows abundance with red as high abundance and blue as low abundance; n = 10–11/group with independent experiments performed in singlicate measurements.

FIGURE 5.

Relative abundance of bacterial taxa at the genus level in male and female offspring exposed to varying micronutrients prenatally. Taxonomy was assigned based on the SILVA 138 database at 99% similarity. Diets are abbreviated as HFNC, high 10x folic acid (20 mg/kg diet) with no choline; HFRC, high 10x folic acid (20 mg/kg diet) with recommended choline (1 g free choline/kg diet); HV, high 10x vitamins; RV: recommended vitamins. n = 10–11/group with independent experiments performed in singlicate measurements.

3.4. Correlation

A correlation matrix was constructed to show the association among the outcome measures with the main goal of drawing meaningful relations within our complex dataset containing significant differences (Figure 6A, B). In both male and female offspring, hypothalamic 5‐HTR2C protein expression was positively correlated with brain 5‐HT concentrations (r = 0.39, p < 0.05 in males; r = 0.38, p < 0.05). The relationship between refeeding response and analytical measures differed, which a negative correlation between refeeding response and 5‐HTR2C expression occurred in male offspring (r = −0.36, p < 0.05) whereas refeeding response was negatively correlated with brain 5‐HT concentrations in female offspring (r = −0.46, p < 0.01). There was a negative correlation between colon 5‐HT and brain 5‐HT in both male and female offspring (r = −0.58, p < 0.0001 in males; r = −0.53, p < 0.01 in females), and colon 5‐HT was also trended to be positively correlated with refeeding response in female offspring (r = 0.29, p = 0.08). For bacterial genera, Lactococcus was negatively correlated with colon 5‐HT concentrations in male offspring (r = −0.41, p < 0.01). Ruminococcus, Bacteroides, and Oscillospira were of importance in female offspring, which Ruminococcus was negatively correlated with brain 5‐HT (r = −0.37, p < 0.05) and there was a trend for a negative correlation with 5‐HTR2C expression (r = −0.29, p = 0.096); Bacteroides was negatively correlated with colon 5‐HT concentrations (r = −0.31, p < 0.05); and Oscillospira was positively correlated with refeeding response (r = 0.31, p < 0.05) and negatively correlated with brain 5‐HT (r = −0.35, p < 0.05). Figure 7 depicts a summary of the relationships between specific microbes and their impact on the serotonergic system.

FIGURE 6.

Correlation matrix of 5‐hydroxytryptamine 2C receptor (5‐HTR2C) protein expression, 1‐h refeeding response, brain and colon 5‐hydroxytryptamine (5‐HT) concentrations and bacterial taxa identified from random forest modeling in male and female offspring (A, B). The color saturation and type denote the direction and strength of the correlation with dark blue representing perfect positive correlation and dark red representing perfect negative correlation. The numbers show the Spearman's correlation coefficient; which p ≤ 0.05 is bolded; 0.05 < p ≤ 0.1 is underlined; n = 8–9/group for 5‐HTR2C, n = 8–12/group for refeeding response, n = 9–12/group for 5‐HT measurements and n = 10–11/group for gut microbiota analyses. Independent experiments were performed in technical duplicate measurements for 5‐HT whereas all others were in singlicate measurements.

FIGURE 7.

Summary depicting the effect of differentially abundant microbes identified in this study on 5‐hydroxytryptamine 2C receptor (5‐HTR2C) protein expression, refeeding response to m‐chlorophenylpiperazine (mCPP) and 5‐hydroxytryptamine (5‐HT) concentrations in the brain and colon for male and female offspring. Within colored boxes, the black arrow indicates a significant relationship assessed by the Spearman's correlation coefficient as p ≤ 0.05; the gray arrow indicates a trend for a relationship as 0.05 < p ≤ 0.1; and the dash indicates no relationship.

4. Discussion

This study was the first to examine the brain and gut serotonergic system in conjunction with the gut microbiota in offspring exposed to varying amounts of micronutrients during pregnancy. The findings from this research support our hypothesis that micronutrients consumed above or below the recommendations lead to a dysregulated serotonergic system associated with gut microbiota alterations in male and female offspring. Our key results are summarized to provide an easily interpretable overview in Figure 8. The use of a big data‐driven approach in this study identified top genera of Lactococcus, Ruminococcus, Bacteroides, and Oscillospira that distinguished group effects with important features associated with host serotonin, which would provide a foundation for dissecting perturbations in the gut–brain axis.

FIGURE 8.

Overview of key study results. Prenatal consumption of excess or imbalanced micronutrients leads to lower 5‐hydroxytryptamine (5‐HT) concentrations in the brain, lower hypothalamic 5‐hydroxytryptamine 2C receptor (5‐HTR2C) protein expression with a general failure in responsiveness to m‐chlorophenylpiperazine as manifested by higher refeeding response concurrently with higher colon 5‐HT concentrations. Variations in the gut microbiota composition were found sex‐dependently with male offspring showing a lower abundance of Lactococcus and female offspring showing higher abundances of Ruminococcus and Oscillospira and lower abundance of Bacteroides, which had significant correlations with measures of the serotonergic system expression and function.

A high intake of multivitamins or folic acid with or without choline resulted in diminished 5‐HTR2C expression in the hypothalamus at the protein level. The most well‐known function of 5‐HTR2C is food intake suppression [40], as previously demonstrated using mice with global mutation or knockout of the 5‐htr2c gene [41]. The reduction in 5‐HTR2C protein expression of 30% in the hypothalamus of male offspring and 50% in female offspring exposed to HV, HFRC, and HFNC compared to RV reflects responsiveness to the maternal gestational diet composition that persisted into adulthood. In our previous study that weaned male offspring to a regular‐fat diet with either the recommended or high amount of folic acid for 29 weeks, high folic acid provided postweaning resulted in lower hypothalamic 5‐htr2c mRNA expression in those born to dams fed high folic acid compared to the control, but we did not see differences with the gestational diet alone [22]. Differences in the study design including the postweaning diet composition and timing of tissue collection, in which this study utilized a high‐fat diet for 12 weeks postweaning, may have yielded variations in effects. This study lacked gene expression measurements but based on our prior findings, the level of protein rather than transcript may have rendered higher sensitivity to prenatal micronutrient exposure. Additionally with no differences in mRNA expression of 5‐htr2c with folic acid alone in our prior work [22], we have yet to pursue alterations in DNA methylation with the promoter region of 5‐htr2c as other select genes that showed transcript level changes were chosen for epigenetic analyses. Studies examining 5‐HTR2C regulation by DNA methylation are lacking thus far, and additional investigations into complementary modes of mechanisms beyond protein expression are required.

We observed a general failure to suppress food intake upon mCPP in male offspring from excess or imbalanced exposure to micronutrients with HV and HFRC males displaying higher refeeding response than HFNC males with respect to the control offspring. We recognize that a pharmacological approach with mCPP being one of the most widely available 5‐HTR2C agonists [42] may be useful in revealing the receptor function in the regulation of food intake. Using the Spearman correlation coefficient, we found a negative correlation between hypothalamic 5‐HTR2C protein expression and refeeding response in male offspring, indicating that the faded mCPP action may be connected to a lower abundance of the receptor protein. The female offspring of HFRC‐fed dams showed food intake suppression upon mCPP at a greater extent that differed from HV and HFNC females, without any correlation between 5‐HTR2C expression and refeeding response. Variations in response may arise as the anorexigenic action of the receptor agonist can be inhibited through distinct intracellular signaling in hypothalamic neurons [43], which emphasize the presence of multiple contributors to serotonergic effects. Other subtypes of 5‐HTR and downstream components thus should be considered as modulators of complex behavior [44].

Brain 5‐HT concentrations were lower in male and female offspring that displayed higher colon 5‐HT compared to the control group indicating an overall shift in the central and peripheral pools. Among the total 5‐HT content, the expected percentage of the central pool is estimated to be 5% whereas the peripheral pool derived from the colon is suggested to be over 90% [5], and this relative proportion was observed in male and female offspring from the control group. Male and female offspring of HV and HFNC dams and male offspring of HFRC dams had lower brain 5‐HT and higher colon 5‐HT compared to RV, which coincide with our previously reported obesogenic phenotypes [20, 21, 22, 23]. These findings support another study showing that the level of constitutive 5‐HT tone either favors or provides protection against obesity using rat sublines [45]. Consistent with this, the regulatory role of the central and peripheral 5‐HT network has recently been highlighted as a contributor to various feeding signals [46], denoting the intricacy of the serotonergic pathways that modify higher‐order functions. In contrast, female offspring of HFRC‐fed dams did not show differences in brain nor colon 5‐HT concentrations, which reflect their resistance to the obesogenic phenotypes found in our previous studies [21, 23, 47]. However, correlation analyses revealed that refeeding response was negatively correlated with brain 5‐HT, with a trend for positive correlation with colon 5‐HT in female offspring, suggesting the importance of brain 5‐HT availability in serotonergic functional behavior. With the reduction of 5‐HTR2C protein expression in the hypothalamus but higher response to mCPP in intake suppression in HFRC females, 5‐HT concentrations in the brain and colon may be compensatory suggesting altered regulation depending on diet and sex of offspring.

This study revealed beta diversity differences as assessed by the Jaccard index distance with 45% of the variation in the data explained by the first coordinate axis in the PCoA. We leveraged random forest modeling that uses multiple decision trees to identify bacterial taxa with predictive power for classification by diet, which Adlercreutzia, Lactococcus, Ruminococcus (from Ruminococcaceae), Bacteroides, Lactobacillus, and Oscillospira emerged. These genera have a range of characteristics known to break down nutrient components or produce short‐chain fatty acids with the potential to impact metabolic health [48, 49]. When expressed as relative abundance percentage separated by diet and sex of offspring, subtle differences occurred including elevated Bacteroides, lower Oscillospira in HFRC males, lower Lactococcus and Lactobacillus in HFNC males, and lower Adlercreutzia in HFNC females as well as lower Bacteroides but higher Lactococcus and Lactobacillus in HV males and higher Adlercreutzia in HV females. The current results align with our previous study using different bioinformatics tools, which found that variations in the gut microbiota composition arise with excess or imbalances in micronutrients in male and female offspring [21].

Our correlation analyses were novel in relating gut bacterial taxa with 5‐HTR2C, refeeding response, and 5‐HT, overall demonstrating gut microbiota‐associated serotonergic dysregulation. In male offspring (summarized in Figure 7), higher colon 5‐HT was correlated with lower abundance of Lactococcus, which is a type of lactic acid‐producing bacteria with the potential for lessening the metabolic consequences of a Western‐style diet as demonstrated with Lactococcus lactis subsp. cremoris [50]. Our results are consistent with another study that conversely reported lower colon 5‐HT concentrations using a genetically engineered strain of L. lactis secreting antiinflammatory cytokine interleukin‐10 with protective effects on low‐grade colon inflammation [51]. In female offspring, distinct signatures of correlation were observed with Ruminococcus, Bacteroides, and Oscillospira (summarized in Figure 7). Lower brain 5‐HT was correlated with higher abundance of Ruminococcus belonging to the Ruminococcaceae family from the Firmicutes phylum, as well as a trend for a relationship between lower 5‐HTR2C protein expression and higher Ruminococcus in female offspring. Although not directly comparable, a porcine study reported that soy formula feeding leads to higher Ruminococcus relative to sow feeding with a reduction in colon 5‐HT [52]. Together with our results, various nutritional challenges may influence the availability of 5‐HT coinciding with the impact on Ruminococcus. We also found that higher colon 5‐HT in female offspring was correlated with lower abundance of Bacteroides at the genus level. It is well‐established that Bacteroides is a common dominant genus in the human gut microbiota as nonspore‐forming bacteria with extensive capacity for carbohydrate fermentation [53]. In a different study, B. fragilis, B. uniformis, Segmented Filamentous Bacteria, altered Schaedler flora, and a consortium of Bacteroides species including B. thetaiotaomicron, B. acidifaciens, and B. vulgatus did not impact host colon 5‐HT but spore‐forming bacteria were modulators [8]. Our study utilized fecal samples from the distal colon of Wistar rats with intact microbiota, which may have captured a complex mixture of nonspore‐forming microbes in contrast to mono‐colonization of select Bacteroides. The constraints of 16S rRNA sequencing in the current study inhibited species‐level assignments due to limited variations in the short amplicon sequence but future work may incorporate environment‐weighted taxonomy classifiers for improved resolution [54]. We view that species of Bacteroides outside the predefined sets as well as other nonspore‐forming species or communities may be responsible for reduced peripheral 5‐HT. For Oscillospira, there was a positive correlation with refeeding response, in addition to lower brain 5‐HT being correlated with higher microbial abundance in female offspring. Oscillospira is widely present in animal and human intestines that has been shown to be increased with high‐fat diet‐induced obesity but lower abundance has also been linked with other diseases such as ulcerative colitis and inflammatory bowel disease [55]. Given the emerging evidence of Oscillospira as part of the stable microbiota with its link to a wide range of diseases [56], the capacity to govern host health and behavior may be through the brain serotonergic system.

The significance of our study is that the gut microbiota may serve as a potential target to fine‐tune host 5‐HT concentrations in the central and peripheral compartments with functional impact on the receptor system. We note that the supplemental dose of 10x employed in this study was intended to capture the physiologically relevant intake level found in North America, which the current dietary environment reflects increased consumption of nutritional supplements, including folic acid, and fortified products [16, 17], with a large proportion of pregnant women exceeding the upper tolerable intake level for folic acid [18, 57–60]. However, we acknowledge that although 10x or close to 10x have been reported in other studies [17, 18], along with the recent concerns of unintended consequences arising from excess intake patterns [61, 62], our results need to be confirmed with a lower dose range that surrounds the average population intake level (e.g., 3–5x) for translational value. Thus, the lack of dose response study presents a limitation in our work, which the potential discrepancies between the experimental conditions and real‐world supplementation practices require reconciliation. Another limitation is that our measurements were restricted to 5‐HT concentrations and 5‐HTR2C expression and function, which were selected based on the relevance to the obesogenic phenotypes in our prior reports [21, 23]. Other receptor subtypes as well as transporters and precursors that regulate 5‐HT metabolism, in addition to the transcript level of candidate genes, require further examination to determine comprehensive changes. Further, random forest modeling and correlation analyses focused on the top genera and their relationship with the serotonergic system without elucidating causality. Other factors that influence 5‐HT concentrations require investigation, including short‐chain fatty acids derived from gut microbes [63]. Our previous study has shown sex‐dependent alterations in acetic acid and butyric acid in male and female offspring exposed to varying micronutrients prenatally [21], but analyses of complete bacterial metabolites would be needed to establish a link with the serotonergic system. Finally, 5‐HT signaling is known to be involved in multiple brain regions that regulate homeostatic and hedonic neurocircuitry including the hippocampus [64]. Serotonergic neuronal activity in various areas of the brain should be of interest for mapping functional organization and regulatory mechanisms in relation to alterations of the gut microbiota.

In conclusion, excess consumption of multivitamins or folic acid with or without choline during pregnancy leads to gut microbiota‐associated disruptions in central and peripheral serotonin in offspring. Future studies that connect early life nutrition with the serotonergic system in shaping metabolic disease risk may provide potential therapeutic applications.

Conflicts of Interest

The authors declare no conflicts of interest.

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1002/mnfr.70044

Supporting information

Supporting Information

Dong J., Al‐Issa M., Feeney J. S., Shelp G. V., Poole E. M., Cho C. E., Prenatal Intake of High Multivitamins or Folic Acid With or Without Choline Contributes to Gut Microbiota‐Associated Dysregulation of Serotonin in Offspring. Mol. Nutr. Food Res. 2025, 69, e70044. 10.1002/mnfr.70044

Data Availability Statement

All relevant data are contained within the article or Supporting Information. Sequencing data have been deposited to the NCBI Sequence Read Archive under the accession number of PRJNA1116821.